Self-Organizing 3D Human Neural Tissue Derived from Induced Pluripotent Stem Cells Recapitulate Alzheimer’s Disease Phenotypes

A drug walks into a phase three clinical trial. It cleared amyloid plaques in mice. It protected neurons in mice. It did essentially nothing in humans. The trial is halted. Billions of dollars, years of work, and — most importantly — patients who had nothing else to try. That scene has played out over and over again in Alzheimer's disease research, more than in almost any other field of medicine. Raja and colleagues put the diagnosis plainly: the clinical pipeline has a dismal success rate, and a large part of the explanation is the models we’ve been using to build it. The question their two thousand sixteen paper set out to answer is whether we can grow human brain tissue in a dish that actually gets Alzheimer's. The problem with existing models runs deeper than it might seem. Transgenic mice only recapitulate Alzheimer's pathology to an extent, and there are serious concerns about their validity. Part of the issue is genetic: many of the variants that raise human Alzheimer's risk sit in non-coding regions of the genome that aren’t well conserved between species. This means the regulatory machinery driving human disease is simply absent in a mouse. And even if you get past the animal models and go to human cells in a dish, two-dimensional cultures miss something essential — they lack an interstitial compartment, the physical space between cells where extracellular amyloid aggregates actually form. You can’t see that pathology if there’s no space for it to happen. Raja and colleagues argue that these mismatches — genetic, structural, and dimensional — are what leave Alzheimer's drug development trapped in this cycle of preclinical success and clinical failure. Their solution is the brain organoid: a three-dimensional, self-organizing mass of human neural tissue grown from induced pluripotent stem cells, or iPSCs. Induced pluripotent stem cells are adult cells — typically skin cells or fibroblasts — that have been reprogrammed back into a stem-cell-like state and can then be directed to become neurons and other neural cell types. Critically, they carry the patient's own genome. So if you take fibroblasts from someone with familial Alzheimer's disease and reprogram them into induced pluripotent stem cells, the resulting neurons carry the same disease-causing mutations that person was born with. No genetic engineering needed. No artificial overexpression of amyloid. Just human cells with human disease genetics. Familial Alzheimer's disease, as distinct from the more common late-onset sporadic form, is caused by inherited mutations in a small number of genes — primarily the amyloid precursor protein gene, APP, or the presenilin genes PSEN1 and PSEN2, which are part of an enzyme complex that processes APP. The patient lines Raja and colleagues used carried either a duplication of the APP gene or a mutation in PSEN1. There were multiple different lines, with multiple different mutations. The team adapted a scaffold-free protocol developed by Kadoshima and colleagues to grow these induced pluripotent stem cell-derived cells into dense, self-organizing three-dimensional organoids — no external gel or matrix holding them together, just the cells organizing themselves. Within a month, translucent neuroectodermal regions appeared. By 60 to 90 days, the organoids had developed rosette-like neuroprogenitor zones and dense fields of mature neurons. Because the organoids lack blood vessels, deeper regions tend to become necrotic. The authors measured where apoptotic markers appeared and set a conservative analysis boundary: the outer 250 micrometers from the organoid surface where neurons were healthy and caspase-3 staining was sparse. Everything reported comes from that viable outer band. Now, the key question: do these organoids actually develop Alzheimer's disease? The answer, across three different pathological hallmarks, is yes. Start with amyloid. By enzyme-linked immunosorbent assay — an assay measuring protein concentrations in the culture fluid — organoids from an APP duplication line secreted significantly more amyloid-beta 1-40 and 1-42 than control organoids. Amyloid-beta 42 is the longer, stickier, and more toxic form of the peptide, and it was significantly elevated with a p-value of 0.004. Immunohistochemistry revealed both intracellular and extracellular amyloid immunoreactivity, and particle analysis showed a progressive, time-dependent increase in the number and size of amyloid aggregates at 60 days and again at 90 days. These aggregates grew in both frequency and size as the organoids aged in culture. That age-dependence is important — it mirrors what happens in human disease, where pathology accumulates over time. Then tau. Tau is a protein that normally stabilizes the internal scaffolding of neurons. In Alzheimer's disease, it becomes hyperphosphorylated — too many phosphate groups attached — which causes it to detach, clump together, and form neurofibrillary tangles. At 60 days, there was no significant difference in hyperphosphorylated tau between familial Alzheimer's disease and control organoids. By 90 days, that had changed dramatically. Phosphorylation at two sites — serine 396 and threonine 181 — was significantly elevated in the familial Alzheimer's disease organoids, with p-values of 0.001 and 0.03 respectively. Thioflavin-S staining, which lights up beta-sheet aggregates and is the structural signature of amyloid and tangle pathology, covered a greater area and showed more positive particles in the familial Alzheimer's disease organoids at 90 days. The temporal sequence matters: amyloid pathology appeared first, tau pathology followed. That ordering is consistent with the amyloid cascade hypothesis, the leading framework for how Alzheimer's unfolds. The third hallmark is endosome abnormalities. Endosomes are small intracellular compartments involved in protein trafficking and processing — including, crucially, the processing of APP into amyloid-beta. Early endosome antigen 1 labeling showed a significant increase in large endosomes in the familial Alzheimer's disease organoids, and a functional transferrin endocytosis assay demonstrated that the average size of transferrin-positive endosomes was significantly larger in the disease lines, with a p-value of 0.005. Endosome enlargement is considered an early marker of cellular dysfunction in Alzheimer's, appearing before plaques and tangles become prominent, and here it was showing up in the dish. All three of these phenotypes were reproduced across multiple familial Alzheimer's disease lines. The APP duplication line wasn't the only one showing pathology — an additional APP duplication line and a PSEN1 A264E mutation line both showed increased amyloid aggregates and elevated hyperphosphorylated tau at 90 days. One PSEN1 line, M146I, showed non-significant trends toward higher amyloid and no tau difference, but the group analyses of variance across all lines were highly significant — for amyloid particle counts, p-values of 0.002 and 0.0005 depending on particle size; for tau intensity, a p-value of 0.0002 with an R-squared of 0.66. Different mutations, different induced pluripotent stem cell lines, same pathological picture. That consistency is the model's strongest argument. Then came the drug experiments. The organoids weren’t just a disease-observation platform — they were a test platform. The team treated thirty-day-old familial Alzheimer's disease organoids with two inhibitors: a beta-secretase inhibitor, which blocks beta-secretase, and Compound E, which blocks gamma-secretase. These are the two enzymes that cleave APP to produce amyloid-beta — and blocking them is the same strategy that has been pursued in human clinical trials. Compounds were added starting at day 30 and replenished with every media change, assessed at 60 and 90 days, at two dose levels. After 30 days of treatment, amyloid aggregates were significantly reduced in a dose-dependent manner. After 60 days of treatment, the reduction was even more pronounced. Tau pathology told a more nuanced story: after 30 days of treatment, hyperphosphorylated tau was not significantly affected. But after 60 days of treatment, as assessed at day 90, tau was significantly reduced as well — again dose-dependently, with an F-statistic of 19.82 and a p-value of 0.0001. The sequence is telling. Lower amyloid first, lower tau later. That's the amyloid cascade playing out in reverse when you intervene at the right point, and it supports the idea that tau hyperphosphorylation in this model is downstream of amyloid accumulation. The limits of the model deserve honesty. Without vasculature, the interior becomes necrotic, restricting analysis to that outer 250-micrometer band. And induced pluripotent stem cell reprogramming resets the cellular epigenome toward a more juvenile state, so these organoids don’t carry the accumulated decades of aging that characterize the human brain in late life. Raja and colleagues are explicit: the extent to which the organoid tissue represents the aged human brain has not been fully examined. These are ninety-day-old structures, not seventy-year-old brains. But here's what they are. They are human neural tissue, derived from patients with Alzheimer's disease, carrying disease-relevant mutations, that spontaneously develops amyloid aggregation, hyperphosphorylated tau, and endosome abnormalities in an age-dependent sequence — and responds to the same drug classes that have been tested in humans. That is a qualitatively different kind of experimental tool than a transgenic mouse or a flat dish of neurons. Whether it will predict clinical success better than what came before is an open question. But it at least asks the disease in the right language. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.